OpenAI has solved Navier–Stokes, a massive unsolved problem in mathematics. That's a pretty big deal — and there's already a controversy about it, to do with the mathematicians involved.

But I'm not that interested in the result. What's more striking to me is that a new OpenAI model did this — one that's only been training since August 28, and is already significantly better than Astra.

The new OpenAI internal model vs GPT-6 Astra on a set of open maths problems

The incredible thing is the pace. Only in July were problems like this starting to fall, and this is a major one. Now they're saying the new model is significantly better again.

This is definitely getting towards superintelligence, for sure. Even OpenAI's own chief scientist now says their chain-of-thought monitoring is getting less reliable as the models get more capable.